Windows Developer MCP Server
# Windows Developer MCP Server
[](https://glama.ai/mcp/servers)
[](https://github.com/peeyushcodes/windows-developer-mcp/actions)
[](LICENSE)
[](https://python.org)
[](https://github.com/jlowin/fastmcp)
[](https://github.com/astral-sh/ruff)
[](https://github.com/detachhead/basedpyright)
**Windows Developer MCP** gives AI assistants (**Claude Desktop**, **LM Studio**, **Cursor**, etc.) safe, native developer superpowers on Windows โ including terminal execution, Git/GitHub integration, Docker management, filesystem operations, and AI project diagnostics.
---
## โก 1-Minute Quick Start
### Step 1: Install `uv`
If you don't have `uv` installed, open PowerShell and run:
```powershell
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
```
### Step 2: Configure Your AI Client
#### ๐ค Claude Desktop Setup
Edit `%APPDATA%\Claude\claude_desktop_config.json`:
```json
{
"mcpServers": {
"windows-developer-mcp": {
"command": "uv",
"args": [
"--directory",
"C:\\path\\to\\windows-developer-mcp",
"run",
"windows-developer-mcp"
],
"env": {
"MCP_SERVER__PROFILE": "developer",
"MCP_WORKSPACE__PATH": "C:\\path\\to\\your\\project"
}
}
}
}
```
#### ๐ป LM Studio / Local LLM Setup
Edit your LM Studio MCP server configuration:
```json
{
"mcpServers": {
"windows-developer-mcp": {
"command": "uv",
"args": [
"--directory",
"C:\\path\\to\\windows-developer-mcp",
"run",
"windows-developer-mcp"
],
"env": {
"MCP_SERVER__PROFILE": "minimal",
"MCP_WORKSPACE__PATH": "C:\\path\\to\\your\\project"
}
}
}
}
```
---
## ๐ฅ Key Features
- **๐ก๏ธ Workspace-Confined & Sandboxed**: All commands are executed securely with strict path normalization, denylists, and audit logging.
- **โก Dual-Profile Engine**:
- **`minimal`**: Optimized for **Local LLMs (7B/8B)** in LM Studio (<1.2K token context footprint).
- **`developer`**: Ideal for **Claude Desktop / GPT-4o** (~3.8K token footprint).
- **`full`**: Complete exposure for **70B+ LLMs & Glama Indexing**.
- **๐งฐ 40+ Developer Tools across 12 Modules**: Terminal, Git, GitHub, Docker, Filesystem, Python, Node, SQLite, Windows, Network, Browser, and AI Project Analysis.
---
## ๐งฐ Available Tool Modules
| Module | Core Tools Exposed | Description |
| :--- | :--- | :--- |
| **Terminal** | `terminal_run`, `terminal_get_session` | Safe PowerShell & CMD subprocess execution. |
| **Filesystem** | `filesystem_read`, `filesystem_write`, `filesystem_list` | Workspace-restricted file & directory operations. |
| **Git** | `git_status`, `git_log`, `git_diff`, `git_commit`, `git_push` | Git branch, commit, diff, and remote management. |
| **GitHub** | `github_auth_status`, `github_repo_info`, `github_list_issues`, `github_list_prs` | GitHub API for repos, PRs, issues, and releases. |
| **Docker** | `docker_list_containers`, `docker_logs`, `docker_run`, `docker_build` | Container, image, and Compose orchestration. |
| **Python** | `python_run`, `python_check_package`, `python_install_package` | Virtual environment management & execution. |
| **Node.js** | `node_run`, `npm_run`, `npm_install` | Node.js script execution & npm package management. |
| **SQLite** | `sqlite_query`, `sqlite_schema`, `sqlite_list_tables`, `sqlite_execute` | Database queries & schema inspection. |
| **Windows** | `windows_system_info`, `windows_list_processes`, `windows_get_env` | System metrics, process list, and environment vars. |
| **Network** | `network_ping`, `network_dns_lookup`, `network_port_scan` | Network diagnostics & connectivity checks. |
| **Project (AI)** | `project_analyze`, `project_security_scan`, `project_generate_readme` | Tech stack detection, security audit, & code gen. |
| **Browser** | `browser_open_url`, `browser_fetch_page`, `browser_extract_text` | Web page fetching & clean text extraction. |
---
## ๐ง AI-Native Project Analysis Tools
`ProjectProvider` provides intelligent tools designed specifically for codebase comprehension:
- ๐ **`project_analyze`**: Auto-detects frameworks, tech stacks, and entry points.
- ๐ก๏ธ **`project_security_scan`**: Static audit for hardcoded API keys, tokens, and credentials.
- ๐ณ **`project_summarize`**: Hierarchical directory tree generator.
- ๐ฆ **`project_dependencies`**: Cross-language dependency analysis.
- ๐ **`project_generate_readme`**: Drafts a complete README.md based on codebase analysis.
- ๐งช **`project_generate_tests`**: Generates pytest unit test skeletons for source files.
---
## ๐ก๏ธ Safety & Security Features
1. **Workspace Boundary Sandbox**: Operations are strictly restricted to the specified workspace path (`MCP_WORKSPACE__PATH`).
2. **Command Denylist**: Blocks dangerous operations (`format`, `shutdown`, `del /f /s /q`, etc.).
3. **Audit Logging**: All executed actions are recorded in `logs/audit.log`.
---
## ๐ Documentation Suite
- ๐ [Architecture Overview](docs/architecture.md)
- ๐ [Provider Reference](docs/providers.md)
- ๐ก๏ธ [Security Model](docs/security.md)
- ๐ค [LM Studio Setup Guide](docs/lm_studio_setup.md)
- ๐ฅ๏ธ [Claude Desktop Setup Guide](docs/claude_desktop_setup.md)
- ๐ค [Contributing Guide](docs/contributing.md)
- ๐ [Glama Manifest (glama.json)](glama.json)
- ๐ [Changelog](CHANGELOG.md)
---
## ๐ License
Distributed under the **MIT License**. See [LICENSE](LICENSE) for details.
TDQS
Scored across 112 tools
Most tools are separated by clear domain prefixes such as docker_, git_, github_, sqlite_, and npm_, which helps differentiate them. However, several tools overlap in purpose: fetch_page/extract_text/http_get/check_url all deal with web content, run_powershell and run_script both execute PowerShell scripts, and tree/list_directory/search_files share file-discovery territory.
The dominant pattern is snake_case verb_noun or domain_verb_noun naming, e.g. delete_file, git_commit, docker_stop, and npm_install. Minor deviations like tree, ping, traceroute, and github_auth_status break the pattern slightly, but overall the naming is predictable and not chaotic.
112 tools is an extreme count for a single MCP server, spanning file operations, shell execution, Docker, system information, Python, Node.js, Git, GitHub, networking, SQLite, and project analysis. This broad sprawl imposes a heavy selection burden on agents and goes far beyond a well-scoped tool set.
The server covers a wide range of developer workflows, but individual clusters have notable gaps: GitHub only lists/gets issues and PRs without create/merge/update operations, Docker lacks image removal and container start, and environment variables can be read but not set. These holes force agents to fall back to generic shell commands or external workflows.